Assessing Critical Adaptations in Automated Adaptive Software Systems by Stage Decomposition

Abstract

In environments involving a variety of connected devices and systems, there is an ever-increasing demand for automated adaptation. To ensure that all threats are identified and manageable in such environments, quality assurance activities including testing and inspections in design-time should focus on assessing the reliability of critical adaptations, which may threaten life, economic property, or important information. This paper proposes an approach for identifying and evaluating critical adaptations on the basis of their automation level, reliability, detectability, and recoverability by decomposing adaptations into four stages: monitor, analyze, plan, and execute. This paper also empirically evaluates the effectiveness of the proposed approach by assessing a real safety-critical telecommunication system with critical adaptation features and comparing the results with the STAMP (System Theoretic Accident Model and Processes)/STPA (System-Theoretic Process Analysis) approach. The results of the evaluation indicated that the proposed approach could assess critical adaptation features provided by the system with reasonable effort. Additionally, structured views provided by the proposed approach enable efficient quality assurance activities. In the evaluation, the proposed approach achieves similar results to the STAMP/STPA approach but requires 33% less effort.